The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires the installation of the ticketfairy package via NPM and connects to an MCP server at https://www.ticketfairy.com/docs/mcp to retrieve documentation and tools. These resources are provided by the official vendor.
[COMMAND_EXECUTION]: The agent is instructed to use the ticketfairy command-line interface to perform financial operations, such as listing brands and checking balances. The instructions emphasize using the --read-only flag unless changes are explicitly requested and approved by the user.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external guides and search results via MCP tools (search_guides, get_guide). This creates an attack surface where instructions embedded in the guides could influence agent behavior. * Ingestion points: MCP tool outputs from search_guides and get_guide (SKILL.md). * Boundary markers: Explicit instructions to "Read it before you act", "Follow those steps; do not guess", and "Keep secrets out of the chat" (SKILL.md). * Capability inventory: Execution of ticketfairy CLI commands and generation of prompts for the "Fai" dashboard assistant (SKILL.md). * Sanitization: Instructions to use specific brand identifiers and follow the exact steps provided in the guides rather than interpreting freely (SKILL.md).

use-vault-and-capital